lems related to questions of sampling, variability, multiple comparisons, and appropriateness of a given test for group differences, which were not discussed at any length, have rendered the results of some studies questionable. Too, the question of the reliability of measuring instruments used has not been given a degree of attention in the literature which could be called adequate. Notable, also, by its absence has been any concern with the measurement of abstract performance at lower levels of mental deficiency. This development may be in part a function of the level of difficulty of instructions and other requirements for performance in the traditional abstract-concrete tasks.
For future research purposes it would be constructive to conclude this section with a brief discussion of the major questions to be considered in evaluating any presumed dimension of individual differences. Initially, one would raise the question of intraindividual generality; i.e., how consistent is an S's performance from one abstraction task to another. For example, do Ss who perform well in sorting tasks also give a high proportion of verbal abstractions in vocabulary definitions and take fewer trials to criterion in a discrimination learning situation in which a response is being strengthened to a common property of dissimilar objects? Assuming adequate reliability, if an S's relative position in a distribution changes markedly from one abstraction task to another, the generality (and thus the importance) of the postulated dimension would be seriously questioned. Given a certain degree of generality, one might then raise the question of whether the proposed dimension should be considered as an independent factor or as a component of a broader construct, such as intelligence. Thirdly, an attempt would be made to identify the range of environmental conditions under which group differences held. The familiar experimental by subject variable design would be of immeasurable value in evaluating interactions. One would want to know, for example, whether differences between normals and mental defectives hold for all levels of task difficulty. Lastly, an intensive experimental study would be carried out of the antecedent environmental variables which facilitate and interfere with the performance of mental defectives on abstraction tasks.
In a sense, we have been outlining a long-range research program designed to evaluate the validity of the construct, abstract versus concrete behavior. Unfortunately, there appear to have been no sustained attempts in the area of mental deficiency to consider any of the questions raised here. It is impossible to avoid the impression that research to date has been essentially an attempt to demonstrate differences which some writers have concluded are reliably present in the populations under analysis. The paucity of experimental research in this area should suggest further the strength of preconceptions concerning the durability of postulated group differences.
The present review leads to disappointment, in terms of not only the limited number and variety of investigations, but also the neglect of problems of statistical design, terminology, and the logic of formal theo-